the effect of education on health: evidence from national … · 2020-03-26 ·...
TRANSCRIPT
SERIEs (2020) 11:83–103https://doi.org/10.1007/s13209-019-0201-0
ORIG INAL ART ICLE
The effect of education on health: evidence from nationalcompulsory schooling reforms
Raquel Fonseca1,2 · Pierre-Carl Michaud2,3,4 · Yuhui Zheng5
Received: 9 October 2018 / Accepted: 18 July 2019 / Published online: 1 August 2019© The Author(s) 2019
AbstractThis paper sheds light on the causal relationship between education and health out-comes. We combine three surveys (SHARE, HRS and ELSA) that include nationallyrepresentative samples of people aged 50 and over from fourteen OECD countries. Weuse variation in the timing of educational reforms across these countries as an instru-ment for education. Using IV-probit models, we find causal evidence that more yearsof education lead to better health. One additional year of schooling is associated with6.85 percentage points (pp) reduction in reporting poor health and 3.8 pp and 4.6 ppreduction in having self-reported difficulties with activities of daily living (ADLs) andinstrumental ADLs, respectively. The marginal effect of education on the probabilityof having a chronic illness is a 4.4 pp reduction. This ranges from a reduction of 3.4pp for heart disease to a 7 pp reduction for arthritis. The effects are larger than thosefrom a probit model that does not control for the endogeneity of education. However,we do not find conclusive evidence that education reduces the risk of cancer, strokeand psychiatric illness.
Keywords Education · Health · Causality · Compulsory schooling laws
JEL Classification I1 · I2
1 Introduction
There is abundant evidence on the relationship between education and health.1 Manystudies, whether country-specific (Etile 2014; Kim 2016) or international, have docu-
1 For a review, see Grossman (2005).
We thank Simon Lord for his excellent research assistance. This research was supported by the NationalInstitute on Aging, under the grant R01 AG040176-06 and P01AG008291. Errors are our own.
B Raquel [email protected]
Extended author information available on the last page of the article
123
84 SERIEs (2020) 11:83–103
mented that this relationship exists in multiple countries, although magnitudes mightdiffer (Banks et al. 2006; Andreyeva et al. 2007; Mackenbach et al. 2008; Michaudet al. 2011). These results deserve attention: if the link is causal, then the effect ofeducation on health should be taken into account when designing education and healthpolicies. This paper aims to estimate the causal effect of education on various healthoutcomes using cross-country variation in the timing of education reforms.
Many mechanisms have been suggested on how education could improve health:raising efficiency in health production (productive efficiency) (Grossman 1972),changing inputs in health production (allocative efficiency) (Grossman 2005), chang-ing time preference (Becker and Mulligan 1997), changing behavioral patterns, e.g.,smoking, obesity, preventive care (Huisman et al. 2005; Mackenbach et al. 2008;Barcellos et al. 2018); and finally, gainingmore resources, e.g., higher income, occupa-tional status, better housing, better food, better quality of care, and living environment(i.e., Case and Deaton 2005; Cutler and Lleras-Muney 2008).
However, there is a persistent problem with these studies. Observational studiesexamining correlations between education and health cannot be interpreted causallybecause education might be endogenous. First, earlier health endowments could affectboth education and health later in life. Second, an unobserved variable, like timepreference, genetic factors or family background, could affect both education andhealth. To solve this problem, we follow the line in the literature which has employedinstitutional changes as instruments for education, for instance, Lleras-Muney (2005),Cutler and Lleras-Muney (2006), Cutler and Lleras-Muney (2008), Clark and Royer(2013), Brunello et al. (2016) and Galama et al. (2018), among others.2 This literaturehas generated different results. Our goal is to add to the literature on understandingthe causal effect of education on health.
We use three data sets including nationally representative samples of individualsaged 50 and over from 14OECD countries. These are the Health and Retirement Study(HRS) for the USA, the English Longitudinal Study of Ageing (ELSA), and the Studyof Health, Ageing and Retirement in Europe (SHARE). We instrument educationusing differences in educational reform across these countries, the hypothesis beingthat different compulsory schooling laws can affect education differently across birthcohorts and across countries in an exogenous way, given that the laws can change bytime and/or by country. We consider compulsory schooling laws that would impactindividuals born between 1905 and 1955. For eight out of the fourteen countries in ouranalysis, a nationwide change in compulsory schooling laws was noted for cohortsborn between those years. For the other countries, there was either no such change, orthe change varied geographically within a country.
We find that more years of education lead to a lower probability of self-reportingpoor health (SRH) and self-reported difficulties with activities of daily living (ADLs)as well as instrumental ADLs (IADLs), and lower prevalence in chronic illness. Morespecifically, one additional year of schooling is associated with 6.85 percentage points(pp) reduction in reporting poor health, 4.6 and 3.8 pp reduction in having ADL andIADL limitations. Concerning the chronic illness indicator, we document a 4.4 ppreduction: 2.7 pp reduction for diabetes, 3.3 pp reduction for heart disease, 4.6 pp
2 See next section for the details.
123
SERIEs (2020) 11:83–103 85
reduction for hypertension, 7 pp reduction for arthritis and 1.4 pp for lung disease.These effects are larger and different than the probit estimates alone, which do notcontrol for the endogeneity of education. We do not find conclusive evidence thateducation has a causal effect on cancer, stroke and psychiatric illness.
This paper adds to the current literature in the following ways. First, we find acausal relationship between education and health by using cross-country variation incompulsory schooling laws over time as an instrumental variable. Second, we examinea wide range of health outcomes, from SRH to functional status and instrumentalfunctional status, and a set of chronic conditions. This differentiates ourwork from thatof recent literature, for instance, Galama et al. (2018), who focus mainly on mortalityand its two most common preventable behavioral causes, smoking and obesity, orGathmann et al. (2015), who also focuses on mortality, or even Crespo et al. (2014),Brunello et al. (2013) andMazzonna (2014). This also differentiates ourwork from thatof Brunello et al. (2016), who concentrate on self-reported health and health behaviors(smoking, drinking, exercising and the body mass index) for European countries.
2 Literature review
In this section we review the subset of the literature which is the most relevant to ourstudy.
One study using US data instrumented education using individual quarter of birthand family background (Adams 2002). It found a positive effect of education on self-reported health (SRH hereafter) and functional status in both OLS and IV estimates.Another study (Lleras-Muney 2005) explored state variation in compulsory educationlaws in the USA as instruments for education and found that each additional yearof education lowers the probability of dying in the next 10 years by as much as3.6 percentage points. Mazumder (2008), extending Lleras-Muney’s analysis usingdifferent data, found even larger effects of education on health. Oreopoulos (2006),using compulsory schooling law changes as an instrument for education as well, alsofound a statistically significant relationship between education and SRH in the UK, inaddition to a negative effect of education on physical andmental disability in the USA,while Silles (2009) found that increased schooling caused more self-reported goodhealth and lowered probabilities of long-term illness, activity-limiting experience, andwork-preventing experience. Fischer et al. (2013) also found evidence, for Sweden,suggesting that education reduces mortality. Brunello et al. (2016) also found thateducation caused more self-reported good health and affects health behaviors for a setof European countries.
Those results are compelling, butmany studies also reportmixed results. Jürges et al.(2013), for instance, finds no causal effects between education and the two biomarkersunder study. Furthermore, the effect of education on SRH was positively significantonly among older female cohorts, whereas it was negative among younger of womenand insignificant among men regardless of age. Kemptner et al. (2010), instrumentingyears of schooling using variations in the timing of the introduction of a 9th gradein West Germany, found that more schooling caused less long-term illness, less workdisability and less obesity among men but not women. Smoking behavior was not
123
86 SERIEs (2020) 11:83–103
causally affected by education in either gender. More mixed evidence is reported byvan Kippersluis et al. (2011), who use Dutch data. Their estimates show that an extrayear of schooling reduces the probability of dying for older men. Similarly, Gathmannet al. (2015) find that more education yields small mortality reductions in the shortand long run for men, while women, in contrast, appear to experience no mortalityreductions from compulsory schooling reforms. More recently, Galama et al. (2018)reviewed the evidence on education’s causal effects on smoking and obesity. Theyfound no convincing evidence of an effect of education on obesity, while the effectson smoking are not apparent when duration of schooling is increased. They becomesignificant only when schooling reforms affect individual schooling track or their peergroup. An effect of education on mortality does appear to exist in some contexts, butnot in others, and seems to depend on gender, on labor market returns of education,and whether education affects the quality of an individual’s peers.
Many more studies simply fail to find an association between education and health.Albouy and Lequien (2009), for example, who also instrument education using schoolreform, fail to find a statistically significant causal effect of education on mortal-ity in France. Similarly for Britain, Clark and Royer (2013) find little evidence thatadditional schooling improves health outcomes or changes health behaviors. In Den-mark, Arendt (2005) found that the IV estimates of education on SRH and body massindex (BMI) were statistically insignificant and not statistically different from thoseestimated using OLS.
In short, studies examining the causality between education and health have gen-erated different results. Focusing on older populations across different countries, ourgoal is to shed further light on the causal effect of education on a broad set of healthoutcomes. Doing so, our study is perhaps closest to that of Brunello et al. (2016),who use compulsory laws as instruments in SHARE and ELSA focusing on healthbehaviors between men and women.
3 Data and descriptive analysis
3.1 HRS, SHARE and ELSA
We focus on individuals aged 50 and over in fourteen countries using comparable sur-vey data: the USA, England, and twelve continental European countries.3 Our maindata sources are the three longitudinal surveys on aging: the Health and RetirementStudy (HRS) in the USA, the English Longitudinal Study of Ageing (ELSA) in Eng-land, and the Study of Health, Ageing and Retirement in Europe (SHARE). Thesesurveys were specifically designed to be comparable with one another and each tar-geted people living in the community and aged 50 and over. Follow-up surveys wereconducted biennially. We used data from wave 10 of HRS (2006), wave 3 of ELSA(2006), and wave 2 of SHARE (2006), all of which had been collected between 2006and 2007.
3 Countries included in our study fromSHAREare:Austria, Sweden, theNetherlands, Italy, France,Greece,Switzerland, Belgium, Germany, Czech Republic, Poland and Spain.
123
SERIEs (2020) 11:83–103 87
All study surveys contain a large set of measures of health status, most of whichare comparable across surveys. We constructed from the data a set of subjective andobjective measures of health outcomes. Subjective measures for this analysis includedoverall self-rated health status (SRH), self-reported difficulties with activities of dailyliving (ADLs) as well as instrumental ADLs (IADLs). SRH was measured by askingrespondents to rate their health on a five-point scale: excellent, very good, good,fair, poor. We defined a binary variable of “poor health”, which takes value of 1 ifthe self-rated health is fair or poor and 0 otherwise. Transforming these variablesin dichotomous ones preserve homogeneity of methods in terms of binary choicemodels. We check our results with alternative SRH cutoffs. For limitations in ADLs,questions were asked in all surveys about difficulties in five basic activities: bathing,dressing, eating, getting in and out of bed, andwalking across a room. Individuals wereclassified as having anyADL limitation if they reported limitationswith one ormore ofthe five activities. Limitations in IADLs were measured by questions about difficultiesin the following five activities: making meals, shopping, making phone calls, takingmedications and managing money. Those who reported having some difficulty withany of the five activities were classified as having any IADL limitation.
Objective measures included in all surveys were the same set of doctor-diagnoseddisease questions on cancer, diabetes, hypertension, heart disease, stroke, lung disease,arthritis, and psychological illness.4 We created a binary variable “any chronic” whenindividuals report having any of these chronic conditions. We also analyzed thesevariables one by one.
Our main independent variable is “years of education”. In HRS, respondents wereasked about the highest grade of school or year of college completed. In ELSA, thevariable is defined by the age at which respondents finished full-time education. Weconverted the values into years of education by subtracting the agewhen the respondentleft school by the usual school started age of five. In the second wave of SHARE, therespondents were asked directly about years of full-time education.
Other demographic variables include gender and age. In order to ascertain therobustness of our results, we also consider employment status (working versus nonworking), marital status (1. Married/partner 2. Divorced/separated 3. Widowed and 4.Never married), household size, and mother and father alive.
3.1.1 Descriptive statistics
Table 1 presents summary statistics of the data in more detail. We report the numberof observations (between 51,000 and 53,700, depending on the variables), the meanresponses and standard deviations, and the minimum and maximum values. 34% ofthe sample reported poor health, and 73% of the sample had one or more diagnosedchronic conditions. The prevalence for specific health conditions ranges from 5% forstroke and 44% for hypertension. For functional status, 13% reported having one ormore ADL limitations, while 10% reported having one or more IADL limitations. Our
4 The measure of a “psychiatric illness” in SHARE is different from those in HRS and ELSA. In HRS andELSA there is a question of “Have you ever had or has a doctor ever told you that you had any emotional,nervous, or psychiatric problems?” In SHARE the closest measure is from the question of “Has there beena time or times in your life when you suffered from symptoms of depression which lasted at least 2 weeks?”
123
88 SERIEs (2020) 11:83–103
Table 1 Summary statistics Variables Obs Mean SD Min Max
Health variables
Poor health 53,492 0.34 0.47 0 1
1 + chronic illness 53,634 0.73 0.44 0 1
1 + ADLs 53,517 0.13 0.34 0 1
1 + IADLs 53,516 0.10 0.31 0 1
Cancer 53,467 0.08 0.27 0 1
Diabetes 53,519 0.14 0.35 0 1
Heart disease 53,518 0.18 0.39 0 1
Hypertension 53,534 0.44 0.50 0 1
Arthritis 53,528 0.36 0.48 0 1
Lung disease 53,521 0.07 0.26 0 1
Stroke 53,530 0.05 0.22 0 1
Psychiatric illness 53,399 0.16 0.37 0 1
SES variables
Years of education 52,077 11.00 3.85 0 25
Age 53,634 66.07 9.73 50 89
Cohort 53,634 3.82 1.95 1 8
Male 53,634 0.45 0.50 0 1
Marital status 51,991 1.55 0.92 1 4
Employment status 53,320 0.30 0.46 0 1
Household size 53,634 2.18 1.05 1 14
Family background
Mother alive 51,966 0.20 0.40 0 1
Father alive 52,073 0.08 0.27 0 1
Data source: HRS wave 10, SHARE wave 2, and ELSA wave 3. Dataare weighted by sampling weight (normalized by country)
key independent variable “years of education” ranges from no education to 25 yearsof education with a mean of 11 years with a standard deviation of 3.85. The averageage of the sample is 66 years old ranging from 50 to 89 years, and 45% of respondentsare male.
3.1.2 Health and education
Table 2 shows the unadjusted prevalence of health outcomes by education and bycountry. For ease of presentation, we recode years of education into three categories—tertiary, secondary, and primary or less—based on the educational system in eachcountry. “tertiary” indicates the category with the highest level of education, while“primary” indicates the category with the lowest level of education. In the first column,we list the percentage reporting poor health. In all countries, there is a clear gradientfor the relationship between education and poor health, with those in the lowest levelof education reporting worse health than those in the middle category, while these
123
SERIEs (2020) 11:83–103 89
Table2
Health
outcom
esby
levelsof
educationandby
coun
try.Datasource:HRSwave10
,SHAREwave2,
andELSA
wave3.
Dataareweigh
tedby
samplingweigh
t(normalized
bycoun
try)
High
Medium
Low
High
Medium
Low
High
Medium
Low
High
Medium
Low
England
USA
Austria
Germany
Poor
health
(1)
1625
4316
2852
2030
4828
4160
1+chronicillness
(2)
6569
8180
8891
4557
6657
6370
1+ADLs
(3)
814
2610
1527
69
187
1120
1+ID
Ls
(4)
48
188
1325
46
203
820
Cancer
(5)
87
913
1316
23
25
45
Diabetes
(6)
78
1215
1928
810
158
1320
Heartdisease
(7)
1720
2620
2531
810
1611
1214
Hyp
ertension
(8)
3640
5148
5765
2736
3935
3844
Arthritis
(9)
2533
4651
6268
715
1910
1418
Lun
gdisease
(10)
35
107
1215
24
75
66
Stroke
(11)
24
76
712
32
43
45
Psychiatricillness
(12)
1211
1016
1824
610
1415
1516
Sweden
Netherland
Spain
Italy
Poor
health
(1)
1729
3620
2537
2222
5217
3053
1+chronicillness
(2)
5155
6449
5562
5452
7048
5875
1+ADLs
(3)
36
135
410
24
145
514
1+ID
Ls
(4)
15
105
39
13
115
413
Cancer
(5)
65
65
44
31
25
43
Diabetes
(6)
49
116
713
710
161
814
Heartdisease
(7)
1112
207
1011
33
108
913
Hyp
ertension
(8)
2330
3821
2331
3023
3627
3245
Arthritis
(9)
1010
127
1212
1715
3210
2140
123
90 SERIEs (2020) 11:83–103
Table2
continued
High
Medium
Low
High
Medium
Low
High
Medium
Low
High
Medium
Low
Lun
gdisease
(10)
14
44
58
45
73
39
Stroke
(11)
34
52
35
01
31
24
Psychiatricillness
(12)
1813
1317
2018
1717
2612
1620
France
Greece
Switzerland
Belgium
Poor
health
(1)
1931
459
1634
1214
2620
2537
1+chronicillness
(2)
5965
6940
4966
4652
5660
6167
1+ADLs
(3)
48
143
39
54
108
817
1+ID
Ls
(4)
25
143
28
22
74
413
Cancer
(5)
55
52
22
35
44
34
Diabetes
(6)
49
117
813
85
78
711
Heartdisease
(7)
98
156
714
74
910
1215
Hyp
ertension
(8)
1928
3122
2841
2226
3332
3237
Arthritis
(9)
1624
306
1123
811
1321
2126
Lun
gdisease
(10)
24
61
25
33
34
47
Stroke
(11)
21
42
13
22
42
24
Psychiatricillness
(12)
2823
217
1011
1916
1215
1416
Czech
R.
Poland
Poor
health
(1)
2937
5242
5574
1+chronicillness
(2)
7470
7766
6979
1+ADLs
(3)
67
911
1632
1+ID
Ls
(4)
14
97
926
Cancer
(5)
44
54
32
Diabetes
(6)
1113
1610
914
123
SERIEs (2020) 11:83–103 91
Table2
continued
High
Medium
Low
High
Medium
Low
High
Medium
Low
High
Medium
Low
Heartdisease
(7)
1011
1720
1625
Hyp
ertension
(8)
4043
4542
3848
Arthritis
(9)
1014
1920
2941
Lun
gdisease
(10)
24
66
46
Stroke
(11)
64
59
47
Psychiatricillness
(12)
4134
3320
2120
Datasource:H
RSwave10
,SHAREwave2,andELSA
wave3.
Dataareweigh
tedby
samplingweigh
t(no
rmalized
bycoun
try)
123
92 SERIEs (2020) 11:83–103
latter report worse health than those in the highest category. We are aware that self-reported health could be subject to different measurement errors (see Jürges 2007).In our empirical analysis, we will address this by controlling for countries specificeffects.
The second row shows the percentage of people with any chronic condition. Ameri-cans report higher levels of chronic disease than Europeans. England, Germany, Spain,Italy, France, Poland, Czech Republic and Belgium report higher levels of chronic dis-ease than the other European countries. Countries with larger gradients in reportingconditions by education are the USA, England, Greece, Italy, Austria, Spain, Polandand Czech Republic and Germany.
Rows 3 and 4 show the percentages of people with difficulties for I(ADLs). Allcountries present very high proportions of low-educated individuals with I(ADLs)difficulties compared to the middle-educated and high-educated ones. In particular,these differences are larger in England, the USA, and Poland.
In rows 5 to 12, we show the prevalence of specific health conditions. The threemost prevalent conditions are hypertension, arthritis and heart disease. Americans andEnglish report higher percentages of cancer, arthritis and lung disease than othercountries in each education category. Americans, Polish and Czech report higherpercentages of diabetes in each educational category than the rest of the countries.For hypertension, arthritis, heart disease and lung disease, low-educated individualsreported higher percentages than high-educated ones in every country. For diabetes, thepercentages are higher in the low-educated relative to the high-educated in twelve outof the fourteen countries, with the exception of Switzerland, inwhich the low-educatedhave a prevalence of 7%while the high-educated have a prevalence of 8%. For cancer,the prevalence is higher among the high-educated, relative to that among the low-educated, for four out of the fourteen countries. For stroke, the prevalence is higherfor the low-educated group relative to that of the high-educated group in all countrieswith the exception of Poland, where the prevalence of the low-educated is 7%while theprevalence of the high-educated is 9%. Finally, the patterns of self-reported psychiatricillness by education differ by country. In England, Sweden, France, Switzerland andthe Czech Republic, the higher-educated individuals report more psychiatric illnessthan the rest of the countries where the proportions are closer between different levelsof education.
Table 3 shows the correlations between health and years of education withoutadjusting for other variables. Our results are in line with the literature. There is anegative correlation between poor health and years of education. All the correlationcoefficients are statistically significant at the 1% level. The more educated are lesslikely to report poor health, any chronic condition, or any ADL or IADL limitations.One exception is the positive correlation of ever diagnosedwith cancer with education.This positive correlation for cancer is consistent with other studies (e.g., Smith 2004;Cutler and Lleras-Muney 2008). Possible explanations are that more educated peopleare more likely to visit the doctors and are diagnosed earlier, survive longer, or havespecific risk factors related to years of education, like late childbearing amongwomen.The correlation between self-reported health and years of education is stronger thanthe relationships between education and other outcomes.
123
SERIEs (2020) 11:83–103 93
Table 3 Correlation coefficients of years of education and health outcomes
Poor health 1 + Chronic illness 1 + ADLs 1 + IADLs Cancer Diabetes
−0.248 −0.040 −0.114 −0.122 0.066 −0.056
Heart disease Hypertension Arthritis Lung disease Stroke Psychiatric illness
−0.027 −0.038 −0.007 −0.047 −0.017 −0.020
All Spearman correlations are significant at 1%. Data source: HRS wave 10, SHARE wave 2, and ELSAwave 3. Data are weighted by sampling weight (normalized by country)
3.2 Compulsory schooling laws and Education
To examine the causal relationship between education and health, we use the cross-country variation in compulsory schooling laws over time as instruments for yearsof education. Our hypothesis is that different compulsory schooling laws can affecteducation differently across birth cohorts and across countries in an exogenous way,given that the laws can change by time and/or by country. Since individuals in oursample are aged 50 years and older, we consider compulsory schooling laws thatwould impact individuals born between year 1905 and year 1955. For eight out ofthe fourteen countries in our analysis, there was a nationwide change in compulsoryschooling law for cohorts born between 1905 and 1955. For the other countries therewas either no such law change or the change varied geographically within a country.We obtain the information from different data sources. In England, the 1944 EducationAct raised theminimum school leaving age from 14 to 15, for cohorts born inApr 1933or later (Oreopoulos 2006; Jürges et al. 2013; Silles 2009); In France, minimum schoolleaving age was raised from 13 to 14 for those born after 1923 (Albouy and Lequien2009); Information on compulsory schooling and reforms for Austria, Greece, Italy,Netherlands and Sweden was obtained from a paper by Murtin and Viarengo (2007).Finally, in Czech Republic the compulsory schooling age was 8 years from 1869.Withthe education reform in 1948, the compulsory schooling age was increased to 9 years(Filer et al. 1999). Five other countries are assumed to have no compulsory schoolinglaw changes that would affect the birth cohorts from 1905 to 1955. For Switzerland,Belgium and Spain, the documented compulsory education reforms took place in 1970or later and did not affect cohorts in our sample. The compulsory education reform forPoland was in the 1960’s, only affecting a small subset of our sample. For Germanyand USA, education reforms varied across geographic areas within the country andwe are not able to define a beginning date for a nationwide reform in compulsoryschooling law.
Table 4 reports the average years of education by country, for those aged 50 and overusing our sample. The table also shows the years of compulsory attendance requiredbefore and after compulsory schooling law changes for each country, as well as thefirst birth cohort that were subject to compulsory schooling law changes. The averageyears of education for aged 50 and over are lowest in Spain (7.41 years), and highestin the USA (12.85 years).
123
94 SERIEs (2020) 11:83–103
Table4
Average
yearsof
educationandminim
umyearsof
educationrequ
ired
before
andaftercompu
lsoryscho
olinglawchanges,by
coun
try
Cou
ntry
AVG.Y
earsof
education
Minim
umyearsof
educationrequ
ired
before
compu
lsory
Scho
olinglawchange
Firstb
irth
cohortaffected
bycompu
lsory
scho
olinglawchange
Minim
umyearsof
educationrequired
after
compu
lsoryscho
oling
lawchange
Austria
8.58
819
499
Eng
land
10.80
919
3310
Sweden
11.19
819
369
Netherland
11.05
619
388
Italy
7.92
519
528
France
11.18
719
238
Greece
8.36
619
529
Czech
R.
12.21
819
359
Poland
9.08
7Nonatio
nwidecompu
lsoryscho
olinglawchange
Switzerland
11.20
8Nonatio
nwidecompu
lsoryscho
olinglawchange
Belgium
11.67
8Nonatio
nwidecompu
lsoryscho
olinglawchange
Germany
12.29
8Nonatio
nwidecompu
lsoryscho
olinglawchange
Spain
7.41
7Nonatio
nwidecompu
lsoryscho
olinglawchange
USA
12.85
8.42
Nonatio
nwidecompu
lsoryscho
olinglawchange
Average
yearsof
educationarefrom
HRS,
ELSA
andSH
AREweigh
teddata(normalized
bycoun
try).Informationon
minim
umyearsof
educationrequ
ired
andcompu
lsory
scho
olinglawchangesismainlyob
tained
from
Murtin
andViareng
o(200
7),w
iththefollo
wingexceptions:for
Britain,the
compu
lsoryscho
olinglawandchangeshave
been
describedin
severalp
apersinclud
ingJürges
(200
9);for
Denmark,
theinform
ationcomes
from
Arend
t(20
05)andMurtin
andViareng
o(200
7);for
France,the
compu
lsory
lawreform
was
describedinAlbou
yandLequien
(200
9);for
Czech
Repub
lic,the
inform
ationcomes
from
Fileretal.(199
9);for
theUSA
,com
pulsoryscho
olinglawsvaried
byStates
andchangedatdifferentpointsintim
efrom
year19
15to19
39,asdescribedinLleras-Mun
ey(200
5).W
ecalculated
theaverageminim
umyearsof
educationacross
allcoun
triesdu
ring
thisperiod
,based
onthedatasetof
“Com
pulsoryAtte
ndance
andChild
Labor
StateLaw
s”,p
rovidedby
Adriana
Lleras-Mun
eyon
herwebsite:http://
www.econ.ucla.edu/alleras/research/data.html
123
SERIEs (2020) 11:83–103 95
Fig. 1 Adjusted proportion of years of education (left) and reporting poor health (right) by birth cohorts.Average proportion of reporting poor health is based a probit regression of self-reported poor health againstgender, birth cohort dummies, country dummies. The sample includes individuals born 5 years before andafter the birth cohort affected by schooling law change. Data source: HRS wave 10, ELSA wave 3, SHAREwave 2. Data weighted by sampling weight
In Fig. 1, we draw the reduced-form relationship between compulsory schoolinglaw changes and both, years of education and one of the health outcomes, i.e., poorhealth. We pool the data from the eight countries with law changes and calculate theproportion of years of education (left graph) and the reporting poor health (right graph)by birth cohort, for individuals born 5 years before and after the first cohort affectedby law changes, adjusting for gender, cohort and country. There is a sharp reductionin the proportion reporting poor health for the cohorts affected after the year of reformand the downward shift persists after that year, increasing slightly after the secondcohort.
4 Empirical strategy
We first model the effect of education on different health outcomes using a probitmodel. Hj,i indicates health outcome j , a binary measure, for an individual i , Hj,i
takes the value of 1 if the underlying latent variable, H∗j,i , is positive and zero otherwise.
Edi represents education for the individual, measured as years of education obtained.Xi contains a set of demographic variables: gender, birth cohort dummies for nine agegroups, and country. One could control for other characteristics such as income andmarital status, but then we run the risk of over-controlling as education reforms mightimpact these outcomes as well which will then impact health. Pooling the three datasets, we estimate the latent variable H∗
j,i for all health outcomes. For example, weestimate the probability that an individual is in “poor health” or the probability thatan individual has any chronic disease using the following model:
H∗j,i = α j + Ediβ j + Xiγ j + ε j
H j,i = I (H∗j,i ≥ 0) (1)
where ε j is a random error that is normally distributed.
123
96 SERIEs (2020) 11:83–103
However, we know that education can be endogenous. As mentioned in the intro-duction, different factors can drive this endogeneity, such as reverse causality orunobserved heterogeneity. This potential endogeneity can be addressed with an instru-mental variable probit model in a two-equation model. The set of equations (2) isequivalent to the estimation (1) above. In equation (3), we model education as a func-tion of a set of control variables, Xi , as well as Zi , which is the minimum years ofeducation required for a given individual and varies by country and birth cohort.
H∗j,i = α j + Ediβ j + Xiγ j + ε j
H j,i = I (H∗j,i ≥ 0) (2)
Edi = θ + Xiλ + Ziφ + v j (3)
ε j , v j are random errors that are normally distributed. The key coefficient of interestis β j . If education is exogenous, the probit estimation of Eq. (1) generates an unbiasedestimation of β j . However, education might be endogenous. We therefore use aninstrumental variable approach to estimate β j . The variable, Zi minimum years ofeducation required, is the instrument. Since we control for gender, country dummies,and birth cohort in both stages of the model, the effect of Zi on Edi is estimated aftertaking into account the country- and cohort-specific effects. Zi should not affect healthoutcomes other than through its effect on years of education. This cannot be directlytested as we only have one instrument. However, it is a reasonable assumption if therewere no co-occurrence factors that affected both compulsory schooling law changesand health. The use of compulsory school law changes frommultiple countries reducesthe possibility of such co-occurrence. Our empirical strategy is very similar to the onethat Lleras-Muney (2005) does for USA. We use the compulsory schooling minimumage laws across different countries, and we analyze the effects of these changes oneducation and different health outcomes. We study the across-country variability inoutcomes by comparing countrieswhere reformswere implemented compared to othercountries where no reform was implemented. These latter countries serve as controlgroups. We cluster the standard errors at the birth year–country level.
We then first estimate a probit model based on Eq. (1) for each health outcome. Forthe instrumental variable approach, we jointly estimate (2) and (3) using maximumlikelihood and assume ε j , v j are multivariate normal with correlation coefficient ρ j .We then test whether ρ j is statistically different from zero to see whether the edu-cation variable is exogenous or if the multivariate probit model estimates based on(1) is appropriate. Probit and IV-probit models are estimated using the “probit” and“ivprobit” commands in Stata.
5 Results
5.1 Health and education across countries
We first estimate model (1) to replicate this evidence across countries and using HRS,ELSA and SHARE data sets. The main results are reported in Table 5 for different
123
SERIEs (2020) 11:83–103 97
Table 5 Probit models of years of education on health outcomes (coefficients reported as marginal effects)
Poor health 1 + Chronic illness 1+ ADLs 1+IADLs
Years of education −0.0285*** −0.0063*** −0.0110*** −0.0098***
(0.0012) (0.0007) (0.0005) (0.0004)
N 51,955 52,077 51,998 51,997
Mean 0.3353 0.7295 0.1338 0.1049
Cancer Diabetes Heart disease Hypertension
Years of education 0.0032*** −0.0064*** −0.0048*** −0.007***
(0.0003) (0.0006) (0.0004) (0.0008)
N 51,952 52,002 52,001 52,017
Mean 0.0803 0.1389 0.1849 0.4414
Arthritis Lung disease Stroke Psychiatric illness
Years of education −0.0098*** −0.0045*** −0.0010*** −0.0020***
(0.0009) (0.0004) (0.0002) (0.0007)
N 52,012 52,005 52,014 51,898
Mean 0.3645 0.0722 0.0522 0.1650
Robust standard errors in parentheses, clustered at birth year–country level. All models control for gender,cohort dummies, and country dummies. Data source: HRS wave 10, SHARE wave 2, and ELSA wave 3.Data are weighted by sampling weight (normalized by country)***p < 0.01, **p < 0.05, *p < 0.1
health outcomes. The table provides the marginal effects of years of education onhealth. The coefficients are all negative and significant at the 1% level, meaning thatmore education is associated with lower probability of having health problems. Theonly exception is cancer, for which the coefficient is positive and significant, as notedin the unadjusted results. For SRH, each additional year of schooling is associatedwith a 3 pp reduction in reporting poor health. Each additional year of schooling isalso associated with a 1.1 pp reduction in having ADL and 1 pp reduction in havingIADL limitations. As for chronic conditions, themarginal effects are smaller and rangefrom 1 pp reduction for arthritis to 0.2 pp reduction for psychiatric illness. There is a0.3 pp increase in the incidence of cancer. All models control for gender, country, andbirth cohort. In addition, country dummies are included in all specifications to accountfor institutional and cultural differences. The complete tables are available from theauthors upon request.
5.1.1 Causal relationship between health and education
We next turn to instrumental variable estimation to examine the effects of educationand health.
The first-stage estimation is a linear regression of the individual’s years of educationagainst minimum years of education required by compulsory schooling laws, control-ling for gender, birth cohort, and country. The estimate is statistically significant at
123
98 SERIEs (2020) 11:83–103
Table 6 First-stage results
Years of education
Minimum years of schooling required 0.5623***
First-stage results for minimum years of schooling required are equal across health outcomes. Robuststandard errors in parentheses, clustered at birth year–country level. All models control for gender, cohortdummies, and country dummies. Data source: HRS wave 10, SHARE wave 2, and ELSA wave 3. Data areweighted by sampling weight (normalized by country)***p < 0.01, **p < 0.05, *p < 0.1
Table 7 IV-Probit models of years of education on health outcomes (coefficients reported as marginaleffects)
Poor health 1 + Chronic illness 1+ ADLs 1+IADLs
Years of education −0.0685*** −0.0442*** −0.0385*** −0.0460***
(0.0038) (0.0073) (0.0067) (0.0075)
N 51,955 52,077 51,998 51,997
Mean 0.3353 0.7295 0.1338 0.1049
Cancer Diabetes Heart disease Hypertension
Years of education 0.0024 −0.0274*** −0.0338*** −0.0464***
(0.0058) (0.0085) (0.0071) (0.0077)
N 51,952 52,002 52,001 52,017
Mean 0.0803 0.1389 0.1849 0.4414
Arthritis Lung disease Stroke Psychiatric illness
Years of education −0.0703*** −0.0139** −0.0070 0.0000
(0.0044) (0.0060) (0.0051) (0.0089)
N 52,012 52,005 52,014 51,898
Mean 0.3645 0.0722 0.0522 0.1650
Robust standard errors in parentheses, clustered at birth year-country level. All models control for gender,cohort dummies, and country dummies. Data source: HRS wave 10, SHARE wave 2, and ELSA wave 3.Data are weighted by sampling weight (normalized by country)***p < 0.01, **p < 0.05, *p < 0.1
the 1% level. Raising minimum years required of education by one year increased theaverage years of education by 0.56 years (around 6.4 months) (see Table 6).
Table 7 shows second-stage estimations for each of the binary health outcomes.Marginal effects and robust standard errors are displayed. The second-stage estimationis a probit model of a health outcome against years of education, gender, birth cohort,and country.The results aremixed for the second-stage estimates. For nineof the twelvehealth outcomes, i.e., poor health, any chronic illness, ADL, IADLs, diabetes, heartdisease, hypertension and arthritis lung disease, the effect of education remains nega-tive. The coefficients are all significant at the 1% level, except for lung disease, whichis significant at the 5% level and cancer, stroke and psychiatric illness which are nolonger significant. The magnitudes of the point estimates are much larger when using
123
SERIEs (2020) 11:83–103 99
IV estimation, as compared to the previous model. For example, the marginal effect ofeducation on the probability of reporting poor health increases from 2.8 to 6.85 pp andboth ADL and iADLs functional status indicators increase from 1 percentage point to3.8 and 4.6 percentage points, respectively, which represent non-negligible effects.
The magnitudes of point estimates are much larger for most of the chronic illnessesusing IV estimation. For example, the marginal effect of education on the probabilityof having any chronic illness increases from 0.6 percentage points to 4.4 percentagepoints, with an average of 72% of individuals with at least one health condition in oursample. In particular, the IVmarginal effects increase inmagnitude compared to probitestimates as follows: for heart disease, from 0.5 percentage points reduction to 3.4 per-centage points, for hypertension from 0.7 percentage points to 4.6 percentage pointsand for arthritis from 1 percentage point to 7 percentage points. Approximately 44%and36%of individuals are ever diagnosedwith hypertension and arthritis, respectively.For the remaining outcomes, themarginal effects of education decrease by 0.7 percent-age points the prevalence of diabetes and by 1.4 percentage points for the prevalenceof lung disease. For the other three outcomes, namely cancer, stroke and psychiatricillness, the coefficients are also larger than using the probit regressions, however, theeffects are no longer significant. For these latter three outcomes, we cannot reject theprobit estimates. Hence, we do not find conclusive evidence that education reducesthe risk of these conditions. For cancer, this can potentially be explained by higherand earlier detection rates among cancer patients (Cutler and Lleras-Muney 2008).
It is not uncommon that IV estimates are larger, probably due to heterogeneoustreatment effects or measurement errors in reported years of education (Card 2001).In the first case, the set of individual who responds to compulsory schooling lawsmight be different and have a different health profile (higher risk) than those who donot respond. Their health outcomes may be more sensitive to an increase in education.
5.2 Robustness
We replicate our analysis with different specifications for age, using age and quadraticage instead of birth cohort dummies.
Given that we use a sample of older respondents, it is possible that part of the causaleffect of education is to improve survival which could lead to selection in terms ofwho survives to be included in the sample. As a robustness check, we perform theanalysis after restricting our sample to individuals in lower age ranges. We restrictthe sample to ages 50–66 and 50–70, and the results are qualitatively similar. We alsocontrol for additional socioeconomic variables as we described in Sect. 3.1, includingemployment status,marital status, and household size. The coefficients are a bit smallerin magnitude but qualitatively similar.
In another set of regressions,we control forwhether parents are alive at the interviewin both the probit and IV-probit models. The rationale is that parent survival reflectsfamily background and genetic factors, which could be correlated with both educationand health. The results for the probit and IV-probit models are qualitatively unchanged.
We perform additional checks. The results are robust to including two instrumentsby adding the minimum compulsory schooling age quadratic. We also perform regres-
123
100 SERIEs (2020) 11:83–103
sions using linear probability models with compulsory schooling laws as instruments.All estimates are again qualitatively unchanged. The results with different cutoffs forself-reported health yield very similar results.
We also extend our analysis to other health behaviors, such as ever smoking, cur-rently smoking and obesity. Probit and OLS regressions for smoking measures show apositive and significant relationship with years of education. For obesity, this rela-tionship was negative. However, the IV-probits and the linear probability modelswith compulsory schooling laws as instrument are no longer statistically significant.Therefore, we cannot make any statement about a causal relationship between healthbehavior and education. This can be due to the mature and elderly sample, e.g., socialsmoking has a different stigma in present times compared to some decades ago, whenboth educated and non-educated people used to smoke. Our results for behavioral out-comes are consistent with cited previous literature that did not find significant evidenceof causality in behavioral outcomes.
Finally, some studies suggest that the effect of education on health could dependon the welfare system of the countries (i.e., Cesarini et al. 2016). We attempted toestimate effects on groups of countries with similar welfare systems to see whethercausal effects of education on health varied. We defined “Northern Europe” to includeSweden and the Netherlands. “Western Europe” includes Austria, Germany, France,and Belgium. “Southern Europe” groups Spain, Italy, and Greece. “Eastern Europe”includes Czech Republic and Poland. The USA and England represent the “Anglo-Saxon” countries. This classification of countries broadly follows prior work, suchas Ferrera (1996), and Esping-Andersen (1990, 1999). We did not find statisticallydifferent causal effects across group of countries and results were generally imprecisedue to the small group of countries by subsets.
6 Conclusion
This paper studies the causal relationship between health outcomes and education. Wecombine three surveys that include nationally representative samples of individualsaged 50 and over from fourteen OECD countries. We use differences in educationalreforms across these countries as an instrument for education. In particular,we considereducational reforms which changed the compulsory number of years of schooling. Wefound that an increase in years of education leads to lower probabilities of reportingpoor health and functional status (ADL and iADLs). The causal relationship betweeneducation and several other chronic conditions, i.e., diabetes, heart disease, hyperten-sion, arthritis, lung disease, and the probability of having at least a chronic illness isstill statistically significant and larger than the probit estimates. Although the iv-probitestimates are larger for all health outcomes compared to the probit estimates, we donot find conclusive evidence that education has a significant effect on cancer, strokeand psychiatric illness.
Magnitudes are substantial. We find that an increase in compulsory schooling yearsof one year reduces the probability of reporting poor health by 6.85 percentage points.Themarginal effects of education on ADL and iADLs functional status are a reductionof 3.8 and, respectively, 4.6 percentage points. The marginal effect of education on the
123
SERIEs (2020) 11:83–103 101
probability of having any chronic illness is a decrease of 4.4 percentage points. Theseeffects suggest that increasing compulsory schooling age had large benefits in termsof health. This would suggest that health benefits from higher education may justifythe cost of interventions aimed at improving the quantity and quality of education incountries that have weaker compulsory schooling laws.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 Interna-tional License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution,and reproduction in any medium, provided you give appropriate credit to the original author(s) and thesource, provide a link to the Creative Commons license, and indicate if changes were made.
References
Adams SJ (2002) Educational attainment and health: evidence from a sample of older adults. Educ Econ10(1):97–109
Albouy V, Lequien L (2009) Does compulsory education lower mortality? J Health Econ 28(1):155–168Andreyeva T, Michaud PC, van Soest A (2007) Obesity and Health in Europeans aged 50 years and older.
Public Health 121:497–509Arendt JN (2005) Does education cause better health? A panel data analysis using school reforms for
identification. Econ Educ Rev 24(2):149–160Banks J, Marmot M, Oldfield Z, Smith JP (2006) Disease and disadvantage in the United States and in
Britain. JAMA 295(17):2037–2045Barcellos S, Carvalho L, Turley P (2018) Education can reduce health disparities related to genetic risk of
obsesity: evidence from a British Reform. Proc Natl Acad Sci (PNAS) 115(42)Becker GS, Mulligan CB (1997) The endogenous determination of time preference. Q J Econ 112(3):729–
758Brunello G, Fabbri D, Fort M (2013) The causal effect of education on body mass: evidence from Europe.
J Labor Econ 31(1):195–223Brunello G, Fort M, Schneeweis N, Winter-Ebmer R (2016) The causal effect of education on health: what
is the role of health behaviors? Health Econ 25(3):314–336Card D (2001) Estimating the return to schooling: progress on some persistent econometric problems.
Econometrica 69(5):1127–1160CaseD, DeatonA (2005) Broken down bywork and sex: how our health declines. In:Wise DA (ed) Analysis
in the economic of aging. The University of Chicago Press, Chicago, pp 185–212Cesarini D, Lindqvist E, Östling R,Wallace B (2016)Wealth, health, and child development: evidence from
administrative data on swedish lottery players. Q J Econ 131(2):687–738Clark D, Royer H (2013) The effect of education on adult health and mortality: evidence from Britain. Am
Econ Rev 103:2087–2120Crespo L, Lopez-Noval B, Mira P (2014) Compulsory schooling, education, depression and memory: new
evidence from SHARELIFE. Econ Educ Rev 43:36–46Cutler DM, Lleras-Muney A (2006) Education and health: evaluating theories and evidence, National
Bureau of Economic Research, Inc, NBER working papers: 12352Cutler DM, Lleras-Muney A, Vogl T (2008) Socioeconomic status and heatlh: dimensions andmechanisms,
National Bureau of Economic Research, Inc, NBER working papers: 14333Esping-Andersen G (1999) Three worlds of welfare capitalism. Social foundations of post- industrial
economies. Policy Press, Oxford University Press,OxfordEsping-Andersen G (1990) Social foundations of postindustrial economies. OUP Oxford, OxfordEtile F (2014) Education policies and health inequalities: evidence from changes in the distribution of Body
Mass Index in France, 1981–2003. Econ Hum Biol 13:46–65Ferrera M (1996) The southern model of welfare in social Europe? J Eur Soc Policy 6(1):17–37Filer RK, Jurajda S, Planovsky J (1999) Education and wages in the Czech and Slovak Republics during
transition. Labour Econ 6(1999):581–593
123
102 SERIEs (2020) 11:83–103
Fischer M, Karlsson M, Nilsson T (2013) Effects of compulsory schooling on mortality: evidence fromSweden. Int J Environ Res Public Health 10(8):3596–3618
Galama T, Lleras-Muney A, van Kippersluis, H (2018) The Effect of education on health and mortality:a review of experimental and quasi-experimental evidence, National Bureau of Economic Research,Inc, NBER working papers: 24225
Gathmann C, Jürges H, Reinhold S (2015) Compulsory schooling reforms, education and mortality intwentieth century Europe. Soc Sci Med 127:74–82
GrossmanM (1972) On the concept of health capital and the demand for health. J Polit Econ 80(2):223–255Grossman M (2005) Education and nonmarket outcomes, National Bureau of Economic Research, Inc,
NBER working papers: 11582HuismanM,Kunst AE,Mackenbach JP (2005) Educational Inequalities in smoking amongmen andwomen
aged 16 years and older in 11 European countries. Tobacco Control 14:106–13Jürges H (2007) True health vs response styles: exploring cross-country differences in self-reported health.
Health Econ 16(2):163–178Jürges H, Kruk E, Reinhold S (2013) The effect of compulsory schooling on health–evidence from biomark-
ers. J Popul Econ 26(2):645–672Kemptner D, Jürges H, Reinhold S (2010) Changes in compulsory schooling and the causal effect of
education on health: evidence from Germany, MEA Working Papers, 200-2010Kim YJ (2016) The long-run effect of education on obesity in the US. Econ Hum Biol 21:100–109Lleras-Muney A (2005) The relationship between education and adult mortality in the United States. Rev
Econo Stud 72(1):189–221Mackenbach JP, Stirbu I, Roskam AJ, Schaap MM, Menvielle G, Leinsalu M, Kunst AE (2008) Socioeco-
nomic inequalities in health in 22 European countries. N Engl J Med 358(23):2468–2481Mazumder B (2008) Does education improve health? A reexamination of the evidence from compulsory
schooling laws. Econ Perspect 33(2):216Mazzonna F (2014) The long lasting effects of education on old age health: evidence of gender differences.
Soc Sci Med 101:129–138MichaudPC,GoldmanD,LakdawallaD,GaileyA,ZhengY (2011)Differences in health between americans
and western Europeans: effects on longevity and public finance. Soc Sci Med 73(2):254–263Murtin F, Viarengo M (2007) The convergence process of compulsory schooling in Western Europe: 1950-
2000. PSE working papersOreopoulos P (2006) Estimating average and local average treatment effects of education when compulsory
schooling laws really matter. Am Econ Rev 96(1):152–175Silles MA (2009) The causal effect of education on health: evidence from the United Kingdom. Econ Educ
Rev 28(1):122–128Smith JA (2004) Unraveling the SES: Health connection. Population and development review. Vol. 30,
supplement: aging. Health, and Public Policy 2004:108–132van Kippersluis H, O’Donnell O, Van Doorslaer E (2011) Long run returns to education. Does schooling
lead to an extended old age? J Hum Resour 46(4):695–721
Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published mapsand institutional affiliations.
123
SERIEs (2020) 11:83–103 103
Affiliations
Raquel Fonseca1,2 · Pierre-Carl Michaud2,3,4 · Yuhui Zheng5
Pierre-Carl [email protected]
Yuhui [email protected]
1 ESG UQAM Montreal Canada Centene Corporation, Los Angeles, USA
2 CIRANO, Montreal, Canada
3 HEC Montreal, Montreal, Canada
4 NBER, Cambridge, USA
5 Centene Corporation, St. Louis, USA
123